What Bailey actually said
Speaking at the annual central-bank gathering, Bailey argued that AI and robotics could help the UK finally escape its productivity rut. His core point was that Britain is on a "distinct growth path" compared with the United States: the labour market is softening, and second-round inflation effects — wage pressure, shop prices — remain subdued. In that environment, productivity gains from automation are exactly what the economy needs to grow faster without reigniting inflation.
The Governor was careful not to promise policy moves. The Bank of England remains focused on returning inflation to its 2% target, and Bailey explicitly avoided committing to any near-term rate change. That is classic central-bank signalling: sound optimistic about the supply side, commit to nothing on the demand side.
There is real substance behind the optimism. AI and robotics are not hypothetical for our own operation at ai-jarvis.eu — we run article pipelines, transcription and text-to-speech services daily on production servers. The productivity gains are concrete. But as anyone running these systems knows, the gap between a central-bank speech and a deployed workflow is where the actual economics live.
The numbers behind the speech
Bailey's rhetoric has to be measured against the UK's actual macro picture, and the numbers are unforgiving:
- 2.9% — July CPI inflation, the highest reading since March, driven largely by energy price shocks linked to international tensions.
- 3.75% — the UK benchmark interest rate, unchanged since December.
- 2% — the official CPI target that the Bank is still chasing.
- 6–3 — the July vote in the Monetary Policy Committee to hold rates steady rather than hike.
Translation: inflation is running above target, energy prices are pushing it up, and the Bank is holding its fire. In this context, Bailey's pitch is essentially an economic bet — that AI-driven productivity can ease the trade-off between growth and inflation. More output per worker, the logic goes, means the economy can run hotter without generating wage spirals.
That is a plausible macro story. It is also, so far, an unproven one in the official statistics. Productivity is notoriously slow to measure, and the UK's post-2008 productivity record is among the weakest in the G7. The Bank is asking Europe to believe that this time, the technology actually delivers.
Why Europe should listen carefully
The UK is no longer in the EU, but the Bailey speech is being watched closely in Brussels and Prague for a simple reason: the same AI deployment pipeline will determine European competitiveness. The timing matters. Since 2 August 2026, the EU AI Act has moved into direct, legally binding enforcement, with the EU AI Office and national authorities empowered to apply transparency rules and systemic-risk compliance demands. The era of voluntary codes of practice and grace periods is over.
That creates a genuine divergence. The UK is charting a lighter-touch regulatory course while the EU enforces a stricter regime. For European companies, the practical consequence is not theoretical: if you sell AI services into the EU, you must comply with the AI Act regardless of where your servers sit — and if you want GDPR-safe, data-resident AI, sovereign or open-weight models increasingly look like the least painful route.
The interesting part is that Bailey's macro-bet and the EU's regulatory reality actually point in the same direction. Both assume that AI becomes embedded in real production systems — the agentic AI workflows doing multi-step tasks with minimal human oversight, not passive chatbots. European startups are already building on open-weight models from Mistral, Llama 4, and GLM-5.3 to run workloads locally and stay inside the regulatory perimeter. That is where the growth Bailey wants will have to come from.
What AI actually costs in 2026
Central bankers see the macro potential. Companies see the invoice. The current market snapshot makes the adoption math clearer than any speech:
- Claude Opus 5 (Anthropic): $5 per million input tokens, $25 per million output tokens — roughly €4.6 / €23 at current rates.
- Gemini 3.7 Flash (Google DeepMind): $0.75 input, $3.75 output — roughly €0.7 / €3.5.
- Grok 4.6 (xAI): $2 input ($0.50 cached) and $6 output for prompts under 200k tokens.
- DeepSeek-V4: standard API rates with a 50% discount for off-peak hours.
- Free open weights: Llama 4, GLM-5.3, Mistral's Shieldstral — zero licensing cost, but you pay for the hardware and the engineer's time.
Do the arithmetic. A European company processing 10 million input tokens and 1 million output tokens per month spends about $75 (~€69) on Claude Opus 5 — but only $11.25 (~€10) on Gemini 3.7 Flash for the same volume. That is a six-fold difference for a workload most production systems wouldn't notice. At scale, the choice between premium and commodity models is the difference between a viable AI pipeline and a line item that gets cut in the next budget review.
The honest takeaway
Bailey is right that AI and robotics are a potential source of faster growth. The mistake would be treating that potential as automatic. Productivity gains require electricity, data quality, integration engineering, and regulatory clarity — and they do not arrive in time for the next inflation print.
This is why the Bank can't wait for a productivity miracle before deciding on rates. The 6–3 hold and 2.9% inflation show the constraint: the Bank's mandate is price stability, not industrial policy. The AI growth story is real, but it is a three-to-five-year story being told at a committee that thinks in quarters.
For European businesses, the takeaway is practical rather than poetic. Benchmark before you buy. Compare cloud APIs against open weights on your own hardware — which is exactly what we publish at AI Arena. And build your AI stack to survive both an inflation cycle and an AI Act audit. The productivity that Bailey promised will be earned in the data centre, not the lecture hall.
Will the Bank of England cut rates soon?
Not according to current signals. The July MPC vote was 6–3 in favour of holding at 3.75%. With CPI at 2.9% — well above the 2% target — and energy shocks fading only slowly, Bailey gave no hint of a near-term cut.
Does the EU AI Act apply to UK companies?
Yes, if they place AI services or models on the EU market. Since the AI Act entered direct enforcement phase on 2 August 2026, the EU AI Office and national authorities can enforce transparency and systemic-risk rules on any provider targeting EU users — including UK-based ones.
Is an open-weight model really cheaper than a cloud API?
At scale, usually yes. Cloud APIs run from $0.75 to $25 per million tokens depending on the model, while open weights (Llama 4, GLM-5.3, Mistral) are free to download. The catch is hardware, electricity, and engineering time — which is why real measurements matter more than marketing claims.
